Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List More from Evan Jones - Software Engineer | Computer Scientist
You can't safely use the C setenv() or unsetenv() functions in a program that uses threads. Those functions modify global state, and can cause other threads calling getenv() to crash. This also causes crashes in other languages that use those C standard library functions, such as Go's os.Setenv (Go issue) and Rust's std::env::set_var() (Rust issue). I ran into this in a Go program, because Go's built-in DNS resolver can call C's getaddrinfo(), which uses environment variables. This cost me 2 days to track down and file the Go bug. Sadly, this problem has been known for decades. For example, an article from January 2017 said: "None of this is new, but we do re-discover it roughly every five years. See you in 2022." This was only one year off! (She wrote an update in October 2023 after I emailed her about my Go bug.) This is a flaw in the POSIX standard, which extends the C Standard to allow modifying environment varibles. The most infuriating part is that many people who could influence the standard or maintain the C libraries don't see this as a problem. The argument is that the specification clearly documents that setenv() cannot be used with threads. Therefore, if someone does this, the crashes are their fault. We should apparently read every function's specification carefully, not use software written by others, and not use threads. These are unrealistic assumptions in modern software. I think we should instead strive to create APIs that are hard to screw up, and evolve as the ecosystem changes. The C language and standard library continue to play an important role at the base of most software. We either need to figure out how to improve it, or we need to figure out how to abandon it. Why is setenv() not thread-safe? The biggest problem is that getenv() returns a char*, with no need for applications to free it later. One thread could be using this pointer when another thread changes the same environment variable using setenv() or unsetenv(). The getenv() function is perfect if environment variables never change. For example, for accessing a process's initial table of environment variables (see the System V ABI: AMD64 Section 3.4.1). It turns out the C Standard only includes getenv(), so according to C, that is exactly how this should work. However, most implementations also follow the POSIX standard (e.g. POSIX.1-2017), which extends C to include functions that modify the environment. This means the current getenv() API is problematic. Even worse, putenv() adds a char* to the set of environment variables. It is explicitly required that if the application modifies the memory after putenv() returns, it modifies the environment variables. This means applications can modify the value passed to putenv() at any time, without any synchronization. FreeBSD used to implement putenv() by copying the value, but it changed it with FreeBSD 7 in 2008, which suggests some programs really do depend on modifying the environment in this fashion (see FreeBSD putenv man page). As a final problem, environ is a NULL-terminated array of pointers (char**) that an application can read and assign to (see definition in POSIX.1-2017). This is how applications can iterate over all environment variables. Accesses to this array are not thread-safe. However, in my experience many fewer applications use this than getenv() and setenv(). However, this does cause some libraries to not maintain the set of environment variables in a thread-safe way, since they directly update this table. Environment variable implementations Implementations need to choose what do do when an application overwrites an existing variable. I looked at glibc, musl, Solaris/Illumos, and FreeBSD/Apple's C standard libraries, and they make the following choices: Never free environment variables (glibc, Solaris/Illumos): Calling setenv() repeatedly is effectively a memory leak. However, once a value is returned from getenv(), it is immutable and can be used by threads safely. Free the environment variables (musl, FreeBSD/Apple): Using the pointer returned by getenv() after another thread calls setenv() can crash. A second problem is ensuring the set of environment variables is updated in a thread-safe fashion. This is what causes crashes in glibc. glibc uses an array to hold pointers to the "NAME=value" strings. It holds a lock in setenv() when changing this array, but not in getenv(). If a thread calling setenv() needs to resize the array of pointers, it copies the values to a new array and frees the previous one. This can cause other threads executing getenv() to crash, since they are now iterating deallocated memory. This is particularly annoying since glibc already leaks environment variables, and holds a lock in setenv(). All it needs to do is hold the lock inside getenv(), and it would no longer crash. This would make getenv() slightly slower. However, getenv() already uses a linear search of the array, so performance does not appear to be a concern. More sophisticated implementations are possible if this is a problem, such as Solaris/Illumos's lock-free implementation. Why do programs use environment variables? Environment variables useful for configuring shared libraries or language runtimes that are included in other programs. This allows users to change the configuration, without program authors needing to explicitly pass the configuration in. One alternative is command line flags, which requires programs to parse them and pass them in to the libraries. Another alternative are configuration files, which then need some other way to disable or configure, to be able to test new configurations. Environment variables are a simple solution. AS a result, many libraries call getenv() (see a partial list below). Since many libraries are configured through environment variables, a program may need to change these variables to configure the libraries it uses. This is common at application startup. This causes programs to need to call setenv(). Given this issue, it seems like libraries should also provide a way to explicitly configure any settings, and avoid using environment variables. We should fix this problem, and we can In my opinion, it is rediculous that this has been a known problem for so long. It has wasted thousands of hours of people's time, either debugging the problems, or debating what to do about it. We know how to fix the problem. First, we can make a thread-safe implementation, like Illumos/Solaris. This has some limitations: it leaks memory in setenv(), and is still unsafe if a program uses putenv() or the environ variable. However, this is an improvement over the current Linux and Apple implementations. The second solution is to add new APIs to get one and get all environment variables that are thread-safe by design, like Microsoft's getenv_s() (see below for the controversy around C11's "Annex K"). My preferred solution would be to do both. This would reduce the chances of hitting this problem for existing programs and libraries, and also provide a path to avoid the problems entirely for new code or languages like Go and Rust. My rough idea would be the following: Add a function to copy one single environment variable to a user-specified buffer, similar to getenv_s(). Add a thread-safe API to iterate over all environment variables, or to copy all variables out. Mark getenv() as deprecated, recommending the new thread-safe getenv() function instead. Mark putenv() as deprecated, recommending setenv() instead. Mark environ as deprecated, recommending environment variable functions instead. Update the implementation of environment varibles to be thread-safe. This requires leaking memory if getenv() is used on a variable, but we can detect if the old functions are used, and only leak memory in that case. This means programs written in other languages will avoid these problems as soon as their runtimes are updated. Update the C and POSIX standards to require the above changes. This would be progress. The getenv_s / C Standard Annex K controversy Microsoft provides getenv_s(), which copies the environment variable into a caller-provided buffer. This is easy to make thread-safe by holding a read lock while copying the variable. After the function returns, future changes to the environment have no effect. This is included in the C11 Standard as Annex K "Bounds Checking Interfaces". The C standard Annexes are optional features. This Annex includes new functions intended to make it harder to make mistakes with buffers that are the wrong size. The first draft of this extension was published in 2003. This is when Microsoft was focusing on "Trustworthy Computing" after a January 2002 memo from Bill Gates. Basically, Windows wasn't designed to be connected to the Internet, and now that it was, people were finding many security problems. Lots of them were caused by buffer handling mistakes. Microsoft developed new versions of a number of problematic functions, and added checks to the Visual C++ compiler to warn about using the old ones. They then attempted to standardize these functions. My understanding is the people responsible for the Unix POSIX standards did not like the design of these functions, so they refused to implement them. For more details, see Field Experience With Annex K published in September 2015, Stack Overflow: Why didn't glibc implement _s functions? updated March 2023, and Rich Felker of musl on both technical and social reasons for not implementing Annex K from February 2019. I haven't looked at the rest of the functions, but having spent way too long looking at getenv(), the general idea of getenv_s() seems like a good idea to me. Standardizing this would help avoid this problem. Incomplete list of common environment variables This is a list of some uses of environment variables from fairly widely used libraries and services. This shows that environment variables are pretty widely used. Cloud Provider Credentials and Services AWS's SDKs for credentials (e.g. AWS_ACCESS_KEY_ID) Google Cloud Application Default Credentials (e.g. GOOGLE_APPLICATION_CREDENTIALS) Microsoft Azure Default Azure Credential (e.g. AZURE_CLIENT_ID) AWS's Lambda serverless product: sets a large number of variables like AWS_REGION, AWS_LAMBDA_FUNCTION_NAME, and credentials like AWS_SECRET_ACCESS_KEY Google Cloud Run serverless product: configuration like PORT, K_SERVICE, K_REVISION Kubernetes service discovery: Defines variables SERVICE_NAME_HOST and SERVICE_NAME_PORT. Third-party C/C++ Libraries OpenTelemetry: Metrics and tracing. Many environment variables like OTEL_SERVICE_NAME and OTEL_RESOURCE_ATTRIBUTES. OpenSSL: many configurable variables like HTTPS_PROXY, OPENSSL_CONF, OPENSSL_ENGINES. BoringSSL: Google's fork of OpenSSL used in Chrome and others. It reads SSLKEYLOGFILE just like OpenSSL for logging TLS keys for debugging. Libcurl: proxies, SSL/TLS configuration and debugging like HTTPS_PROXY, CURL_SSL_BACKEND, CURL_DEBUG. Libpq Postgres client library: connection parameters including credentials like PGHOSTADDR, PGDATABASE, and PGPASSWORD. Rust Standard Library std::thread RUST_MIN_STACK: Calls std::env::var() on the first call to spawn() a new thread. It is cached in a static atomic variable and never read again. See implementation in thread::min_stack(). std::backtrace RUST_LIB_BACKTRACE: Calls std::env::var() on the first call to capture a backtrace. It is cached in a static atomic variable and never read again. See implementation in Backtrace::enabled().
This is a reminder that random load balancing is unevenly distributed. If we distribute a set of items randomly across a set of servers (e.g. by hashing, or by randomly selecting a server), the average number of items on each server is num_items / num_servers. It is easy to assume this means each server has close to the same number of items. However, since we are selecting servers at random, they will have different numbers of items, and the imbalance can be important. For load balancing, a reasonable model is that each server has fixed capacity (e.g. it can serve 3000 requests/second, or store 100 items, etc.). We need to divide the total workload over the servers, so that each server stays below its capacity. This means the number of servers is determined by the most loaded server, not the average. This is a classic balls in bins problem that has been well studied, and there are some interesting theoretical results. However, I wanted some specific numbers, so I wrote a small simulation. The summary is that the imbalance depends on the expected number of items per server (that is, num_items / num_servers). This means workload is more balanced with fewer servers, or with more items. This means that dividing a set of items over more servers makes the distribution more unfair, which is a reason we can get worse than linear scaling of a distributed system. Let's make this more concrete with an example. Let's assume we have a workload of 1000 items, and each server can hold a maximum of 100 items. If we place the exact same number of items on each server, we only need 10 servers, and each of them is completely busy. However, if we place the items randomly, then the median (p50) number of items is 100 items. This means half the servers will have more than 100 items, and will be overloaded. If we want less than a 1% chance of an overloaded server, we need to look at the 99th percentile (p99) server load. We need to use at least 13 servers, which has a p99 load of 97 items. For 14 servers, the average is 77 items, so our servers are on average 23% idle. This shows how the imbalance leads to wasted capacity. This is a bit of an extreme example, because the number of items is small. Let's assume we can make the items 10× smaller, say by dividing them into pieces. Our workload now consists of 10k items, and each server has the capacity to hold 1000 (1k) items. Our perfectly balanced workload still needs 10 servers. With random load balancing, to have a less than 1 in 1000 chance of exceeding our capacity, we only need 11 servers, which has a p99 load of 98 items and a p999 of 100 items. With 11 servers, the average number of items is 910 or 91%, so our servers are only 9% idle. This shows how splitting work into smaller pieces improves load balancing. Another way to look at this is to think about a scaling scenario. Let's go back to our workload of 1000 items, where each server can handle 100 items, and we have 13 servers to ensure we have less than a 1% chance of an overloaded server. Now let's assume the amount of work per item doubles, for example because the service has become more popular, so each item has become larger. Now, each server can hold a maximum of 50 items. If we have perfectly linear scaling, we can double the number of servers from 13 to 26 to handle this workload. However, 26 servers has a p99 of 53 items, so we again have a more than 1% chance of overload. We need to use 28 servers which has a p99 of 50 items. This means we doubled the workload, but had to increase the number of servers from 13 to 28, which is 2.15×. This is sub-linear scaling. As a way to visualize the imbalance, the chart below shows the p99 to average ratio, which is a measure of how imbalanced the system is. If everything is perfectly balanced, the value is 1.0. A value of 2.0 means 1% of servers will have double the number of items of the average server. This shows that the imbalance increases with the number of servers, and increases with fewer items. Power of Two Random Choices Another way to improve load balancing is to have smarter placement. Perfect placement can be hard, but it is often possible to use the "power of two random choices" technique: select two servers at random, and place the item on the least loaded of the two. This makes the distribution much more balanced. For 1000 items and 100 items/server, 11 servers has a p999 of 93 items, so much less than 0.1% chance of overload, compared to needing 14 servers with random load balancing. For the scaling scenario where each server can only handle 50 items, we only need 21 servers to have a p999 of 50 items, compared to 28 servers with random load balancing. The downside of the two choices technique is that each request is now more expensive, since it must query two servers instead of one. However, in many cases where the "item not found" requests are much less expensive than the "item found" requests, this can still be a substantial improvement. For another look at how this improves load balancing, with a nice simulation that includes information delays, see Marc Brooker's blog post. Raw simulation output I will share the code for this simulation later. simulating placing items on servers with random selection iterations=10000 (number of times num_items are placed on num_servers) measures the fraction of items on each server (server_items/num_items) and reports the percentile of all servers in the run P99_AVG_RATIO = p99 / average; approximately the worst server compared to average num_items=1000: num_servers=3 p50=0.33300 p95=0.35800 p99=0.36800 p999=0.37900 AVG=0.33333; P99_AVG_RATIO=1.10400; ITEMS_PER_NODE=333.3 num_servers=5 p50=0.20000 p95=0.22100 p99=0.23000 p999=0.24000 AVG=0.20000; P99_AVG_RATIO=1.15000; ITEMS_PER_NODE=200.0 num_servers=10 p50=0.10000 p95=0.11600 p99=0.12300 p999=0.13100 AVG=0.10000; P99_AVG_RATIO=1.23000; ITEMS_PER_NODE=100.0 num_servers=11 p50=0.09100 p95=0.10600 p99=0.11300 p999=0.12000 AVG=0.09091; P99_AVG_RATIO=1.24300; ITEMS_PER_NODE=90.9 num_servers=12 p50=0.08300 p95=0.09800 p99=0.10400 p999=0.11200 AVG=0.08333; P99_AVG_RATIO=1.24800; ITEMS_PER_NODE=83.3 num_servers=13 p50=0.07700 p95=0.09100 p99=0.09700 p999=0.10400 AVG=0.07692; P99_AVG_RATIO=1.26100; ITEMS_PER_NODE=76.9 num_servers=14 p50=0.07100 p95=0.08500 p99=0.09100 p999=0.09800 AVG=0.07143; P99_AVG_RATIO=1.27400; ITEMS_PER_NODE=71.4 num_servers=25 p50=0.04000 p95=0.05000 p99=0.05500 p999=0.06000 AVG=0.04000; P99_AVG_RATIO=1.37500; ITEMS_PER_NODE=40.0 num_servers=50 p50=0.02000 p95=0.02800 p99=0.03100 p999=0.03500 AVG=0.02000; P99_AVG_RATIO=1.55000; ITEMS_PER_NODE=20.0 num_servers=100 p50=0.01000 p95=0.01500 p99=0.01800 p999=0.02100 AVG=0.01000; P99_AVG_RATIO=1.80000; ITEMS_PER_NODE=10.0 num_servers=1000 p50=0.00100 p95=0.00300 p99=0.00400 p999=0.00500 AVG=0.00100; P99_AVG_RATIO=4.00000; ITEMS_PER_NODE=1.0 num_items=2000: num_servers=3 p50=0.33350 p95=0.35050 p99=0.35850 p999=0.36550 AVG=0.33333; P99_AVG_RATIO=1.07550; ITEMS_PER_NODE=666.7 num_servers=5 p50=0.20000 p95=0.21500 p99=0.22150 p999=0.22850 AVG=0.20000; P99_AVG_RATIO=1.10750; ITEMS_PER_NODE=400.0 num_servers=10 p50=0.10000 p95=0.11100 p99=0.11600 p999=0.12150 AVG=0.10000; P99_AVG_RATIO=1.16000; ITEMS_PER_NODE=200.0 num_servers=11 p50=0.09100 p95=0.10150 p99=0.10650 p999=0.11150 AVG=0.09091; P99_AVG_RATIO=1.17150; ITEMS_PER_NODE=181.8 num_servers=12 p50=0.08350 p95=0.09350 p99=0.09800 p999=0.10300 AVG=0.08333; P99_AVG_RATIO=1.17600; ITEMS_PER_NODE=166.7 num_servers=13 p50=0.07700 p95=0.08700 p99=0.09100 p999=0.09600 AVG=0.07692; P99_AVG_RATIO=1.18300; ITEMS_PER_NODE=153.8 num_servers=14 p50=0.07150 p95=0.08100 p99=0.08500 p999=0.09000 AVG=0.07143; P99_AVG_RATIO=1.19000; ITEMS_PER_NODE=142.9 num_servers=25 p50=0.04000 p95=0.04750 p99=0.05050 p999=0.05450 AVG=0.04000; P99_AVG_RATIO=1.26250; ITEMS_PER_NODE=80.0 num_servers=50 p50=0.02000 p95=0.02550 p99=0.02750 p999=0.03050 AVG=0.02000; P99_AVG_RATIO=1.37500; ITEMS_PER_NODE=40.0 num_servers=100 p50=0.01000 p95=0.01400 p99=0.01550 p999=0.01750 AVG=0.01000; P99_AVG_RATIO=1.55000; ITEMS_PER_NODE=20.0 num_servers=1000 p50=0.00100 p95=0.00250 p99=0.00300 p999=0.00400 AVG=0.00100; P99_AVG_RATIO=3.00000; ITEMS_PER_NODE=2.0 num_items=5000: num_servers=3 p50=0.33340 p95=0.34440 p99=0.34920 p999=0.35400 AVG=0.33333; P99_AVG_RATIO=1.04760; ITEMS_PER_NODE=1666.7 num_servers=5 p50=0.20000 p95=0.20920 p99=0.21320 p999=0.21740 AVG=0.20000; P99_AVG_RATIO=1.06600; ITEMS_PER_NODE=1000.0 num_servers=10 p50=0.10000 p95=0.10700 p99=0.11000 p999=0.11320 AVG=0.10000; P99_AVG_RATIO=1.10000; ITEMS_PER_NODE=500.0 num_servers=11 p50=0.09080 p95=0.09760 p99=0.10040 p999=0.10380 AVG=0.09091; P99_AVG_RATIO=1.10440; ITEMS_PER_NODE=454.5 num_servers=12 p50=0.08340 p95=0.08980 p99=0.09260 p999=0.09580 AVG=0.08333; P99_AVG_RATIO=1.11120; ITEMS_PER_NODE=416.7 num_servers=13 p50=0.07680 p95=0.08320 p99=0.08580 p999=0.08900 AVG=0.07692; P99_AVG_RATIO=1.11540; ITEMS_PER_NODE=384.6 num_servers=14 p50=0.07140 p95=0.07740 p99=0.08000 p999=0.08300 AVG=0.07143; P99_AVG_RATIO=1.12000; ITEMS_PER_NODE=357.1 num_servers=25 p50=0.04000 p95=0.04460 p99=0.04660 p999=0.04880 AVG=0.04000; P99_AVG_RATIO=1.16500; ITEMS_PER_NODE=200.0 num_servers=50 p50=0.02000 p95=0.02340 p99=0.02480 p999=0.02640 AVG=0.02000; P99_AVG_RATIO=1.24000; ITEMS_PER_NODE=100.0 num_servers=100 p50=0.01000 p95=0.01240 p99=0.01340 p999=0.01460 AVG=0.01000; P99_AVG_RATIO=1.34000; ITEMS_PER_NODE=50.0 num_servers=1000 p50=0.00100 p95=0.00180 p99=0.00220 p999=0.00260 AVG=0.00100; P99_AVG_RATIO=2.20000; ITEMS_PER_NODE=5.0 num_items=10000: num_servers=3 p50=0.33330 p95=0.34110 p99=0.34430 p999=0.34820 AVG=0.33333; P99_AVG_RATIO=1.03290; ITEMS_PER_NODE=3333.3 num_servers=5 p50=0.20000 p95=0.20670 p99=0.20950 p999=0.21260 AVG=0.20000; P99_AVG_RATIO=1.04750; ITEMS_PER_NODE=2000.0 num_servers=10 p50=0.10000 p95=0.10500 p99=0.10700 p999=0.10940 AVG=0.10000; P99_AVG_RATIO=1.07000; ITEMS_PER_NODE=1000.0 num_servers=11 p50=0.09090 p95=0.09570 p99=0.09770 p999=0.09990 AVG=0.09091; P99_AVG_RATIO=1.07470; ITEMS_PER_NODE=909.1 num_servers=12 p50=0.08330 p95=0.08790 p99=0.08980 p999=0.09210 AVG=0.08333; P99_AVG_RATIO=1.07760; ITEMS_PER_NODE=833.3 num_servers=13 p50=0.07690 p95=0.08130 p99=0.08320 p999=0.08530 AVG=0.07692; P99_AVG_RATIO=1.08160; ITEMS_PER_NODE=769.2 num_servers=14 p50=0.07140 p95=0.07570 p99=0.07740 p999=0.07950 AVG=0.07143; P99_AVG_RATIO=1.08360; ITEMS_PER_NODE=714.3 num_servers=25 p50=0.04000 p95=0.04330 p99=0.04460 p999=0.04620 AVG=0.04000; P99_AVG_RATIO=1.11500; ITEMS_PER_NODE=400.0 num_servers=50 p50=0.02000 p95=0.02230 p99=0.02330 p999=0.02440 AVG=0.02000; P99_AVG_RATIO=1.16500; ITEMS_PER_NODE=200.0 num_servers=100 p50=0.01000 p95=0.01170 p99=0.01240 p999=0.01320 AVG=0.01000; P99_AVG_RATIO=1.24000; ITEMS_PER_NODE=100.0 num_servers=1000 p50=0.00100 p95=0.00150 p99=0.00180 p999=0.00210 AVG=0.00100; P99_AVG_RATIO=1.80000; ITEMS_PER_NODE=10.0 num_items=100000: num_servers=3 p50=0.33333 p95=0.33579 p99=0.33681 p999=0.33797 AVG=0.33333; P99_AVG_RATIO=1.01043; ITEMS_PER_NODE=33333.3 num_servers=5 p50=0.20000 p95=0.20207 p99=0.20294 p999=0.20393 AVG=0.20000; P99_AVG_RATIO=1.01470; ITEMS_PER_NODE=20000.0 num_servers=10 p50=0.10000 p95=0.10157 p99=0.10222 p999=0.10298 AVG=0.10000; P99_AVG_RATIO=1.02220; ITEMS_PER_NODE=10000.0 num_servers=11 p50=0.09091 p95=0.09241 p99=0.09304 p999=0.09379 AVG=0.09091; P99_AVG_RATIO=1.02344; ITEMS_PER_NODE=9090.9 num_servers=12 p50=0.08334 p95=0.08477 p99=0.08537 p999=0.08602 AVG=0.08333; P99_AVG_RATIO=1.02444; ITEMS_PER_NODE=8333.3 num_servers=13 p50=0.07692 p95=0.07831 p99=0.07888 p999=0.07954 AVG=0.07692; P99_AVG_RATIO=1.02544; ITEMS_PER_NODE=7692.3 num_servers=14 p50=0.07143 p95=0.07277 p99=0.07332 p999=0.07396 AVG=0.07143; P99_AVG_RATIO=1.02648; ITEMS_PER_NODE=7142.9 num_servers=25 p50=0.04000 p95=0.04102 p99=0.04145 p999=0.04193 AVG=0.04000; P99_AVG_RATIO=1.03625; ITEMS_PER_NODE=4000.0 num_servers=50 p50=0.02000 p95=0.02073 p99=0.02103 p999=0.02138 AVG=0.02000; P99_AVG_RATIO=1.05150; ITEMS_PER_NODE=2000.0 num_servers=100 p50=0.01000 p95=0.01052 p99=0.01074 p999=0.01099 AVG=0.01000; P99_AVG_RATIO=1.07400; ITEMS_PER_NODE=1000.0 num_servers=1000 p50=0.00100 p95=0.00117 p99=0.00124 p999=0.00132 AVG=0.00100; P99_AVG_RATIO=1.24000; ITEMS_PER_NODE=100.0 power of two choices num_items=1000: num_servers=3 p50=0.33300 p95=0.33400 p99=0.33500 p999=0.33600 AVG=0.33333; P99_AVG_RATIO=1.00500; ITEMS_PER_NODE=333.3 num_servers=5 p50=0.20000 p95=0.20100 p99=0.20200 p999=0.20300 AVG=0.20000; P99_AVG_RATIO=1.01000; ITEMS_PER_NODE=200.0 num_servers=10 p50=0.10000 p95=0.10100 p99=0.10200 p999=0.10200 AVG=0.10000; P99_AVG_RATIO=1.02000; ITEMS_PER_NODE=100.0 num_servers=11 p50=0.09100 p95=0.09200 p99=0.09300 p999=0.09300 AVG=0.09091; P99_AVG_RATIO=1.02300; ITEMS_PER_NODE=90.9 num_servers=12 p50=0.08300 p95=0.08500 p99=0.08500 p999=0.08600 AVG=0.08333; P99_AVG_RATIO=1.02000; ITEMS_PER_NODE=83.3 num_servers=13 p50=0.07700 p95=0.07800 p99=0.07900 p999=0.07900 AVG=0.07692; P99_AVG_RATIO=1.02700; ITEMS_PER_NODE=76.9 num_servers=14 p50=0.07200 p95=0.07300 p99=0.07300 p999=0.07400 AVG=0.07143; P99_AVG_RATIO=1.02200; ITEMS_PER_NODE=71.4 num_servers=25 p50=0.04000 p95=0.04100 p99=0.04200 p999=0.04200 AVG=0.04000; P99_AVG_RATIO=1.05000; ITEMS_PER_NODE=40.0 num_servers=50 p50=0.02000 p95=0.02100 p99=0.02200 p999=0.02200 AVG=0.02000; P99_AVG_RATIO=1.10000; ITEMS_PER_NODE=20.0 num_servers=100 p50=0.01000 p95=0.01100 p99=0.01200 p999=0.01200 AVG=0.01000; P99_AVG_RATIO=1.20000; ITEMS_PER_NODE=10.0 num_servers=1000 p50=0.00100 p95=0.00200 p99=0.00200 p999=0.00300 AVG=0.00100; P99_AVG_RATIO=2.00000; ITEMS_PER_NODE=1.0 power of two choices num_items=2000: num_servers=3 p50=0.33350 p95=0.33400 p99=0.33400 p999=0.33450 AVG=0.33333; P99_AVG_RATIO=1.00200; ITEMS_PER_NODE=666.7 num_servers=5 p50=0.20000 p95=0.20050 p99=0.20100 p999=0.20150 AVG=0.20000; P99_AVG_RATIO=1.00500; ITEMS_PER_NODE=400.0 num_servers=10 p50=0.10000 p95=0.10050 p99=0.10100 p999=0.10100 AVG=0.10000; P99_AVG_RATIO=1.01000; ITEMS_PER_NODE=200.0 num_servers=11 p50=0.09100 p95=0.09150 p99=0.09200 p999=0.09200 AVG=0.09091; P99_AVG_RATIO=1.01200; ITEMS_PER_NODE=181.8 num_servers=12 p50=0.08350 p95=0.08400 p99=0.08400 p999=0.08450 AVG=0.08333; P99_AVG_RATIO=1.00800; ITEMS_PER_NODE=166.7 num_servers=13 p50=0.07700 p95=0.07750 p99=0.07800 p999=0.07800 AVG=0.07692; P99_AVG_RATIO=1.01400; ITEMS_PER_NODE=153.8 num_servers=14 p50=0.07150 p95=0.07200 p99=0.07250 p999=0.07250 AVG=0.07143; P99_AVG_RATIO=1.01500; ITEMS_PER_NODE=142.9 num_servers=25 p50=0.04000 p95=0.04050 p99=0.04100 p999=0.04100 AVG=0.04000; P99_AVG_RATIO=1.02500; ITEMS_PER_NODE=80.0 num_servers=50 p50=0.02000 p95=0.02050 p99=0.02100 p999=0.02100 AVG=0.02000; P99_AVG_RATIO=1.05000; ITEMS_PER_NODE=40.0 num_servers=100 p50=0.01000 p95=0.01050 p99=0.01100 p999=0.01100 AVG=0.01000; P99_AVG_RATIO=1.10000; ITEMS_PER_NODE=20.0 num_servers=1000 p50=0.00100 p95=0.00150 p99=0.00200 p999=0.00200 AVG=0.00100; P99_AVG_RATIO=2.00000; ITEMS_PER_NODE=2.0 power of two choices num_items=5000: num_servers=3 p50=0.33340 p95=0.33360 p99=0.33360 p999=0.33380 AVG=0.33333; P99_AVG_RATIO=1.00080; ITEMS_PER_NODE=1666.7 num_servers=5 p50=0.20000 p95=0.20020 p99=0.20040 p999=0.20060 AVG=0.20000; P99_AVG_RATIO=1.00200; ITEMS_PER_NODE=1000.0 num_servers=10 p50=0.10000 p95=0.10020 p99=0.10040 p999=0.10040 AVG=0.10000; P99_AVG_RATIO=1.00400; ITEMS_PER_NODE=500.0 num_servers=11 p50=0.09100 p95=0.09120 p99=0.09120 p999=0.09140 AVG=0.09091; P99_AVG_RATIO=1.00320; ITEMS_PER_NODE=454.5 num_servers=12 p50=0.08340 p95=0.08360 p99=0.08360 p999=0.08380 AVG=0.08333; P99_AVG_RATIO=1.00320; ITEMS_PER_NODE=416.7 num_servers=13 p50=0.07700 p95=0.07720 p99=0.07720 p999=0.07740 AVG=0.07692; P99_AVG_RATIO=1.00360; ITEMS_PER_NODE=384.6 num_servers=14 p50=0.07140 p95=0.07160 p99=0.07180 p999=0.07180 AVG=0.07143; P99_AVG_RATIO=1.00520; ITEMS_PER_NODE=357.1 num_servers=25 p50=0.04000 p95=0.04020 p99=0.04040 p999=0.04040 AVG=0.04000; P99_AVG_RATIO=1.01000; ITEMS_PER_NODE=200.0 num_servers=50 p50=0.02000 p95=0.02020 p99=0.02040 p999=0.02040 AVG=0.02000; P99_AVG_RATIO=1.02000; ITEMS_PER_NODE=100.0 num_servers=100 p50=0.01000 p95=0.01020 p99=0.01040 p999=0.01040 AVG=0.01000; P99_AVG_RATIO=1.04000; ITEMS_PER_NODE=50.0 num_servers=1000 p50=0.00100 p95=0.00120 p99=0.00140 p999=0.00140 AVG=0.00100; P99_AVG_RATIO=1.40000; ITEMS_PER_NODE=5.0 power of two choices num_items=10000: num_servers=3 p50=0.33330 p95=0.33340 p99=0.33350 p999=0.33360 AVG=0.33333; P99_AVG_RATIO=1.00050; ITEMS_PER_NODE=3333.3 num_servers=5 p50=0.20000 p95=0.20010 p99=0.20020 p999=0.20030 AVG=0.20000; P99_AVG_RATIO=1.00100; ITEMS_PER_NODE=2000.0 num_servers=10 p50=0.10000 p95=0.10010 p99=0.10020 p999=0.10020 AVG=0.10000; P99_AVG_RATIO=1.00200; ITEMS_PER_NODE=1000.0 num_servers=11 p50=0.09090 p95=0.09100 p99=0.09110 p999=0.09110 AVG=0.09091; P99_AVG_RATIO=1.00210; ITEMS_PER_NODE=909.1 num_servers=12 p50=0.08330 p95=0.08350 p99=0.08350 p999=0.08360 AVG=0.08333; P99_AVG_RATIO=1.00200; ITEMS_PER_NODE=833.3 num_servers=13 p50=0.07690 p95=0.07700 p99=0.07710 p999=0.07720 AVG=0.07692; P99_AVG_RATIO=1.00230; ITEMS_PER_NODE=769.2 num_servers=14 p50=0.07140 p95=0.07160 p99=0.07160 p999=0.07170 AVG=0.07143; P99_AVG_RATIO=1.00240; ITEMS_PER_NODE=714.3 num_servers=25 p50=0.04000 p95=0.04010 p99=0.04020 p999=0.04020 AVG=0.04000; P99_AVG_RATIO=1.00500; ITEMS_PER_NODE=400.0 num_servers=50 p50=0.02000 p95=0.02010 p99=0.02020 p999=0.02020 AVG=0.02000; P99_AVG_RATIO=1.01000; ITEMS_PER_NODE=200.0 num_servers=100 p50=0.01000 p95=0.01010 p99=0.01020 p999=0.01020 AVG=0.01000; P99_AVG_RATIO=1.02000; ITEMS_PER_NODE=100.0 num_servers=1000 p50=0.00100 p95=0.00110 p99=0.00120 p999=0.00120 AVG=0.00100; P99_AVG_RATIO=1.20000; ITEMS_PER_NODE=10.0 power of two choices num_items=100000: num_servers=3 p50=0.33333 p95=0.33334 p99=0.33335 p999=0.33336 AVG=0.33333; P99_AVG_RATIO=1.00005; ITEMS_PER_NODE=33333.3 num_servers=5 p50=0.20000 p95=0.20001 p99=0.20002 p999=0.20003 AVG=0.20000; P99_AVG_RATIO=1.00010; ITEMS_PER_NODE=20000.0 num_servers=10 p50=0.10000 p95=0.10001 p99=0.10002 p999=0.10002 AVG=0.10000; P99_AVG_RATIO=1.00020; ITEMS_PER_NODE=10000.0 num_servers=11 p50=0.09091 p95=0.09092 p99=0.09093 p999=0.09093 AVG=0.09091; P99_AVG_RATIO=1.00023; ITEMS_PER_NODE=9090.9 num_servers=12 p50=0.08333 p95=0.08335 p99=0.08335 p999=0.08336 AVG=0.08333; P99_AVG_RATIO=1.00020; ITEMS_PER_NODE=8333.3 num_servers=13 p50=0.07692 p95=0.07694 p99=0.07694 p999=0.07695 AVG=0.07692; P99_AVG_RATIO=1.00022; ITEMS_PER_NODE=7692.3 num_servers=14 p50=0.07143 p95=0.07144 p99=0.07145 p999=0.07145 AVG=0.07143; P99_AVG_RATIO=1.00030; ITEMS_PER_NODE=7142.9 num_servers=25 p50=0.04000 p95=0.04001 p99=0.04002 p999=0.04002 AVG=0.04000; P99_AVG_RATIO=1.00050; ITEMS_PER_NODE=4000.0 num_servers=50 p50=0.02000 p95=0.02001 p99=0.02002 p999=0.02002 AVG=0.02000; P99_AVG_RATIO=1.00100; ITEMS_PER_NODE=2000.0 num_servers=100 p50=0.01000 p95=0.01001 p99=0.01002 p999=0.01002 AVG=0.01000; P99_AVG_RATIO=1.00200; ITEMS_PER_NODE=1000.0 num_servers=1000 p50=0.00100 p95=0.00101 p99=0.00102 p999=0.00102 AVG=0.00100; P99_AVG_RATIO=1.02000; ITEMS_PER_NODE=100.0
I was wondering: how often do nanosecond timestamps collide on modern systems? The answer is: very often, like 5% of all samples, when reading the clock on all 4 physical cores at the same time. As a result, I think it is unsafe to assume that a raw nanosecond timestamp is a unique identifier. I wrote a small test program to test this. I used Go, which records both the "absolute" time and the "monotonic clock" relative time on each call to time.Now(), so I compared both the relative difference between consecutive timestamps, as well as just the absolute timestamps. As expected, the behavior depends on the system, so I observe very different results on Mac OS X and Linux. On Linux, within a single thread, both the absolute and monotonic times always increase. On my system, the minimum increment was 32 ns. Between threads, approximately 5% of the absolute times were exactly the same as other threads. Even with 2 threads on a 4 core system, approximately 2% of timestamps collided. On Mac OS X: the absolute time has microsecond resolution, so there are an astronomical number of collisions when I repeat this same test. Even within a thread I often observe the monotonic clock not increment. See the test program on Github if you are curious.
The read() and write() system calls take a variable-length byte array as an argument. As a simplified model, the time for the system call should be some constant "per-call" time, plus time directly proportional to the number of bytes in the array. That is, the time for each call should be time = (per_call_minimum_time) + (array_len) × (per_byte_time). With this model, using a larger buffer should increase throughput, asymptotically approaching 1/per_byte_time. I was curious: do real system calls behave this way? What are the ideal buffer sizes for read() and write() if we want to maximize throughput? I decided to do some experiments with blocking I/O. These are not rigorous, and I suspect the results will vary significantly if the hardware and software are different than one the system I tested. The really short answer is that a buffer of 32 KiB is a good starting point on today's systems, and I would want to measure the performance to go beyond that. However, for large writes, performance can increase. On Linux, the simple model holds for small buffers (≤ 4 KiB), but once the program approaches the maximum throughput, the throughput becomes highly variable and in many cases decreases as the buffers get larger. For blocking I/O, approximately 32 KiB is large enough to hit the maximum throughput for read(), but write() throughput improves with buffers up to around 256 KiB - 1 MiB. The reason for the asymmetry is that the Linux kernel will only write less than the entire buffer (a "short write") if there is an error (e.g. a signal causing EINTR). Thus, larger write buffers means the operating system needs to switch to the process less often. On the other head, "short reads", where a read() returns less than the maximum length, become increasingly common as the buffer size increases, which diminishes the benefit. There is a SO_RCVLOWAT socket option to change this that I did not test. The experiments were run on two 16 CPU Google Cloud T2D instances, which use AMD EPYC Milan processors (3rd generation, released in 2021). Each core is a real physical core. I used Ubuntu 23.04 running kernel 6.2.0-1005-gcp. My benchmark program is written in Rust and is available on Github. On localhost, Unix sockets were able to transfer data at approximately 9000 MiB/s. Localhost TCP sockets were a bit slower, around 7000 MiB/s. When using two separate cloud VMs with a networking throughput limit of 32 Gbps = 3800 MiB/s, I needed to use 6 TCP sockets to reliably reach that maximum throughput. A single TCP socket gets around 1400 MiB/s with 256 KiB buffers, with peaks as high as 2200 MiB/s. Experiment 1: /dev/zero and /dev/urandom My first experiment is reading from the /dev/zero and /dev/urandom devices. These are software devices implemented by the kernel, so they should have low overhead and low variability, since other tasks are not involved. Reading from /dev/urandom should be much slower than /dev/zero since the kernel must generate random bytes, rather than just zeros. The chart below shows the throughput for reading from /dev/zero as the buffer size is increased. The results show that the basic linear time per system call model holds until the system reaches maximum throughput (256 kiB buffer = 39000 MiB/s for /dev/zero, or 16 kiB = 410 MiB/s for /dev/urandom). As the buffer size increases further, the throughput decreases as the buffers get too big. This suggests that some other cost for larger buffers starts to outweigh the reduction in number of system calls. Perhaps CPU caches become less effective? The AMD EPYC Milan (3rd gen) CPU I tested on has 32 KiB of L1 data cache and 512 KiB of L2 data cache per core. The performance decreases don't exactly line up with these numbers, but it still seems plausible. The numbers for /dev/urandom are substantially lower, but otherwise similar. I did a linear least-squares fit on the average time per system call, shown in the following chart. If I use all the data, the fit is not good, because the trend changes for larger buffers. However, if I use the data up to the maximum throughput at 256 KiB, the fit is very good, as shown on the chart below. The linear fit models the minimum time per system call as 167 ns, with 0.0235 ns/byte additional time. If we want to use smaller buffers, using a 64 KiB buffer for reading from /dev/zero gets within 95% of the maximum throughput. Experiment 2: Unix and localhost TCP sockets Exchanging data with other processes is the thing I am actually interested in, so I tested Unix and TCP sockets on a single machine. In this case, I varied both the write buffer size and the read buffer size. Unfortunately, these results vary a lot. A more robust comparison would require running each experiment many times, and using some sort of statistical comparison. However, this "quick and dirty" experiment satisfied my curiousity, so I didn't do that. As a result, my conclusions here are vague. The simple model that increasing buffer size should decrease overhead is true, but only until the buffers are about 4 KiB. Above that point, the results start to be highly variable, and it is much harder to draw general conclusion. However, appears that increasing the write buffer size generally is quite helpful up to at least 256 KiB, and often needed as much as 1 MiB to get the highest localhost throughput. I suspect this is because on Linux with blocking sockets, write() will not return until it has written all the data in the buffer, unless there is an error (e.g. EINTR). As a result, passing a large buffer means the kernel can do a lot of the work without needing to switch back to user space. Unfortunately, the same is not true for read(), which often returns "short reads" with any data that is available in the buffer. This starts with buffer sizes around 2 KiB, with the percentage of short reads increasing as the buffer size gets larger. This means the simple model does not hold, because we aren't actually increasing the bytes per read call. I suspect this is a factor which means this microbenchmark is likely not representative of real programs. A real program will do something with the buffer, which will provide time for more data to be buffered in the kernel, and would probably decrease the number of short reads. This likely means larger buffers are in practice more useful than this microbenchmark suggests. As a result of this, the highest throughput often was achievable with small read buffers. I'm somewhat arbitrarily selecting 16 KiB at the best read buffer, and 256 KiB as the best write buffer, although a 1 MiB write buffer seems to be To give a sense of how variable the results are, the plot below shows the local Unix socket throughput for each read and write buffer throughput size. I apologize for the ugly plot. I did not want to spend the time to make it more beautiful. This plot is interactive so you can slice the data to the area of interest. I recommend zooming in to the left hand size with read buffers up to about 300 KiB. The first thing to note is at least on Linux with blocking sockets, the writer will almost never have a "short write", where the write system call returns before writing all the data in the buffer. Unless there is a signal (EINTR) or some other "error" condition, write() will not return until all the bytes are written. The same is not true for reads. The read() system call will often return a "short" read, starting around buffer sizes of 2 KiB. The percentage of short reads generally increases as buffer sizes get bigger, which is logical. Another note is that sockets have in-kernel send and receive buffers. I did not tune these at all. It is possible that better performance is possible by tuning these settings, but that was not my goal. I wanted to know what happens "out of the box" for general-purpose programs without any special tuning. Experiment 3: TCP between two hosts In this experiment, I used two separate hosts connected with 32 Gbps networking in Google Cloud. I first tested the TCP throughput using iperf, to independently verify the network performance. A single TCP connection with iperf is not enough to fully utilize the network. I tried fiddling with some command line options and with Kernel settings like net.ipv4.tcp_rmem and wasn't able to get much better than about 12 Gb/s = 1400 MiB/s. The throughput is also highly varible. Here is some example output with iperf reporting at 2 second intervals, where you can see the throughput ranging from 10 to 19 Gb/s, with an average over the entire interval of 12 Gb/s. To hit the maximum network throughput, I need to use 6 or more parallel TCP connections (iperf -c IP_ADDRESS --time 60 --interval 2 -l 262144 -P 6). Using 3 connections gets around 26 Gb/s, and using 4 or 5 will occasionally hit the maximum, but will also occasionally drop down. Using at least 6 seems to reliably stay at the maximum. Due to this variability, it is hard to draw any conclusions about buffer size. In particular: a single TCP connection is not limited by CPU. The system uses about 40% of a single CPU core, basically all in the kernel. This is more about how the buffer sizes may impact scheduling choices. That said, it is clear that you cannot hit the maximum throughput with a small write buffer. The experiments with 4 KiB write buffers reached approximately 300 MiB/s, while an 8 KiB write buffer was much faster, around 1400 MiB/s. Larger still generally seems better, up to around 256 KiB, which occasionally reached 2200 MiB/s = 17.6 Gb/s. The plot below shows the TCP socket throughput for each read and write buffer size. Again, I apologize for the ugly plot.
This is a post for myself, because I wasted a lot of time understanding this bug, and I want to be able to remember it in the future. I expect close to zero others to be interested. The C standard library function isspace() returns a non-zero value (true) for the six "standard" ASCII white-space characters ('\t', '\n', '\v', '\f', '\r', ' '), and any locale-specific characters. By default, a program starts in the "C" locale, which will only return true for the six ASCII white-space characters. However, if the program changes locales, it can return true for other values. As a result, unless you really understand locales, you should use your own version of this function, or ICU4C's u_isspace() function. An implementation of isspace() for ASCII is one line: /* Returns true for the 6 ASCII white-space characters: \t \n \v \f \r ' '. */ int isspace_ascii(int c) { return c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r' || c == ' '; } I ran into this because On Mac OS X, Postgres switches to the system's default locale, which is something that uses UTF-8 (e.g. en_US.UTF-8, fr_CA.UTF-8, etc). In this case, isspace() returns true for Unicode white-space values, which includes 0x85 = NEL = Next Line, and 0xA0 = NBSP = No-Break Space. This caused a bug in parsing Postgres Hstore values that use Unicode. I have attempted to submit a patch to fix this (mailing list post, commitfest entry). For a program to demonstrate the behaviour on different systems, see isspace_locale on Github.
More in programming
Once you’ve written your strategy’s exploration, the next step is working on its diagnosis. Diagnosis is understanding the constraints and challenges your strategy needs to address. In particular, it’s about doing that understanding while slowing yourself down from deciding how to solve the problem at hand before you know the problem’s nuances and constraints. If you ever find yourself wanting to skip the diagnosis phase–let’s get to the solution already!–then maybe it’s worth acknowledging that every strategy that I’ve seen fail, did so due to a lazy or inaccurate diagnosis. It’s very challenging to fail with a proper diagnosis, and almost impossible to succeed without one. The topics this chapter will cover are: Why diagnosis is the foundation of effective strategy, on which effective policy depends. Conversely, how skipping the diagnosis phase consistently ruins strategies A step-by-step approach to diagnosing your strategy’s circumstances How to incorporate data into your diagnosis effectively, and where to focus on adding data Dealing with controversial elements of your diagnosis, such as pointing out that your own executive is one of the challenges to solve Why it’s more effective to view difficulties as part of the problem to be solved, rather than a blocking issue that prevents making forward progress The near impossibility of an effective diagnosis if you don’t bring humility and self-awareness to the process Into the details we go! This is an exploratory, draft chapter for a book on engineering strategy that I’m brainstorming in #eng-strategy-book. As such, some of the links go to other draft chapters, both published drafts and very early, unpublished drafts. Diagnosis is strategy’s foundation One of the challenges in evaluating strategy is that, after the fact, many effective strategies are so obvious that they’re pretty boring. Similarly, most ineffective strategies are so clearly flawed that their authors look lazy. That’s because, as a strategy is operated, the reality around it becomes clear. When you’re writing your strategy, you don’t know if you can convince your colleagues to adopt a new approach to specifying APIs, but a year later you know very definitively whether it’s possible. Building your strategy’s diagnosis is your attempt to correctly recognize the context that the strategy needs to solve before deciding on the policies to address that context. Done well, the subsequent steps of writing strategy often feel like an afterthought, which is why I think of diagnosis as strategy’s foundation. Where exploration was an evaluation-free activity, diagnosis is all about evaluation. How do teams feel today? Why did that project fail? Why did the last strategy go poorly? What will be the distractions to overcome to make this new strategy successful? That said, not all evaluation is equal. If you state your judgment directly, it’s easy to dispute. An effective diagnosis is hard to argue against, because it’s a web of interconnected observations, facts, and data. Even for folks who dislike your conclusions, the weight of evidence should be hard to shift. Strategy testing, explored in the Refinement section, takes advantage of the reality that it’s easier to diagnose by doing than by speculating. It proposes a recursive diagnosis process until you have real-world evidence that the strategy is working. How to develop your diagnosis Your strategy is almost certain to fail unless you start from an effective diagnosis, but how to build a diagnosis is often left unspecified. That’s because, for most folks, building the diagnosis is indeed a dark art: unspecified, undiscussion, and uncontrollable. I’ve been guilty of this as well, with The Engineering Executive’s Primer’s chapter on strategy staying silent on the details of how to diagnose for your strategy. So, yes, there is some truth to the idea that forming your diagnosis is an emergent, organic process rather than a structured, mechanical one. However, over time I’ve come to adopt a fairly structured approach: Braindump, starting from a blank sheet of paper, write down your best understanding of the circumstances that inform your current strategy. Then set that piece of paper aside for the moment. Summarize exploration on a new piece of paper, review the contents of your exploration. Pull in every piece of diagnosis from similar situations that resonates with you. This is true for both internal and external works! For each diagnosis, tag whether it fits perfectly, or needs to be adjusted for your current circumstances. Then, once again, set the piece of paper aside. Mine for distinct perspectives on yet another blank page, talking to different stakeholders and colleagues who you know are likely to disagree with your early thinking. Your goal is not to agree with this feedback. Instead, it’s to understand their view. The Crux by Richard Rumelt anchors diagnosis in this approach, emphasizing the importance of “testing, adjusting, and changing the frame, or point of view.” Synthesize views into one internally consistent perspective. Sometimes the different perspectives you’ve gathered don’t mesh well. They might well explicitly differ in what they believe the underlying problem is, as is typical in tension between platform and product engineering teams. The goal is to competently represent each of these perspectives in the diagnosis, even the ones you disagree with, so that later on you can evaluate your proposed approach against each of them. When synthesizing feedback goes poorly, it tends to fail in one of two ways. First, the author’s opinion shines through so strongly that it renders the author suspect. Your goal is never to agree with every team’s perspective, just as your diagnosis should typically avoid crowning any perspective as correct: a reader should generally be appraised of the details and unaware of the author. The second common issue is when a group tries to jointly own the synthesis, but create a fractured perspective rather than a unified one. I generally find that having one author who is accountable for representing all views works best to address both of these issues. Test drafts across perspectives. Once you’ve written your initial diagnosis, you want to sit down with the people who you expect to disagree most fervently. Iterate with them until they agree that you’ve accurately captured their perspective. It might be that they disagree with some other view points, but they should be able to agree that others hold those views. They might argue that the data you’ve included doesn’t capture their full reality, in which case you can caveat the data by saying that their team disagrees that it’s a comprehensive lens. Don’t worry about getting the details perfectly right in your initial diagnosis. You’re trying to get the right crumbs to feed into the next phase, strategy refinement. Allowing yourself to be directionally correct, rather than perfectly correct, makes it possible to cover a broad territory quickly. Getting caught up in perfecting details is an easy way to anchor yourself into one perspective prematurely. At this point, I hope you’re starting to predict how I’ll conclude any recipe for strategy creation: if these steps feel overly mechanical to you, adjust them to something that feels more natural and authentic. There’s no perfect way to understand complex problems. That said, if you feel uncertain, or are skeptical of your own track record, I do encourage you to start with the above approach as a launching point. Incorporating data into your diagnosis The strategy for Navigating Private Equity ownership’s diagnosis includes a number of details to help readers understand the status quo. For example the section on headcount growth explains headcount growth, how it compares to the prior year, and providing a mental model for readers to translate engineering headcount into engineering headcount costs: Our Engineering headcount costs have grown by 15% YoY this year, and 18% YoY the prior year. Headcount grew 7% and 9% respectively, with the difference between headcount and headcount costs explained by salary band adjustments (4%), a focus on hiring senior roles (3%), and increased hiring in higher cost geographic regions (1%). If everyone evaluating a strategy shares the same foundational data, then evaluating the strategy becomes vastly simpler. Data is also your mechanism for supporting or critiquing the various views that you’ve gathered when drafting your diagnosis; to an impartial reader, data will speak louder than passion. If you’re confident that a perspective is true, then include a data narrative that supports it. If you believe another perspective is overstated, then include data that the reader will require to come to the same conclusion. Do your best to include data analysis with a link out to the full data, rather than requiring readers to interpret the data themselves while they are reading. As your strategy document travels further, there will be inevitable requests for different cuts of data to help readers understand your thinking, and this is somewhat preventable by linking to your original sources. If much of the data you want doesn’t exist today, that’s a fairly common scenario for strategy work: if the data to make the decision easy already existed, you probably would have already made a decision rather than needing to run a structured thinking process. The next chapter on refining strategy covers a number of tools that are useful for building confidence in low-data environments. Whisper the controversial parts At one time, the company I worked at rolled out a bar raiser program styled after Amazon’s, where there was an interviewer from outside the team that had to approve every hire. I spent some time arguing against adding this additional step as I didn’t understand what we were solving for, and I was surprised at how disinterested management was about knowing if the new process actually improved outcomes. What I didn’t realize until much later was that most of the senior leadership distrusted one of their peers, and had rolled out the bar raiser program solely to create a mechanism to control that manager’s hiring bar when the CTO was disinterested holding that leader accountable. (I also learned that these leaders didn’t care much about implementing this policy, resulting in bar raiser rejections being frequently ignored, but that’s a discussion for the Operations for strategy chapter.) This is a good example of a strategy that does make sense with the full diagnosis, but makes little sense without it, and where stating part of the diagnosis out loud is nearly impossible. Even senior leaders are not generally allowed to write a document that says, “The Director of Product Engineering is a bad hiring manager.” When you’re writing a strategy, you’ll often find yourself trying to choose between two awkward options: Say something awkward or uncomfortable about your company or someone working within it Omit a critical piece of your diagnosis that’s necessary to understand the wider thinking Whenever you encounter this sort of debate, my advice is to find a way to include the diagnosis, but to reframe it into a palatable statement that avoids casting blame too narrowly. I think it’s helpful to discuss a few concrete examples of this, starting with the strategy for navigating private equity, whose diagnosis includes: Based on general practice, it seems likely that our new Private Equity ownership will expect us to reduce R&D headcount costs through a reduction. However, we don’t have any concrete details to make a structured decision on this, and our approach would vary significantly depending on the size of the reduction. There are many things the authors of this strategy likely feel about their state of reality. First, they are probably upset about the fact that their new private equity ownership is likely to eliminate colleagues. Second, they are likely upset that there is no clear plan around what they need to do, so they are stuck preparing for a wide range of potential outcomes. However they feel, they don’t say any of that, they stick to precise, factual statements. For a second example, we can look to the Uber service migration strategy: Within infrastructure engineering, there is a team of four engineers responsible for service provisioning today. While our organization is growing at a similar rate as product engineering, none of that additional headcount is being allocated directly to the team working on service provisioning. We do not anticipate this changing. The team didn’t agree that their headcount should not be growing, but it was the reality they were operating in. They acknowledged their reality as a factual statement, without any additional commentary about that statement. In both of these examples, they found a professional, non-judgmental way to acknowledge the circumstances they were solving. The authors would have preferred that the leaders behind those decisions take explicit accountability for them, but it would have undermined the strategy work had they attempted to do it within their strategy writeup. Excluding critical parts of your diagnosis makes your strategies particularly hard to evaluate, copy or recreate. Find a way to say things politely to make the strategy effective. As always, strategies are much more about realities than ideals. Reframe blockers as part of diagnosis When I work on strategy with early-career leaders, an idea that comes up a lot is that an identified problem means that strategy is not possible. For example, they might argue that doing strategy work is impossible at their current company because the executive team changes their mind too often. That core insight is almost certainly true, but it’s much more powerful to reframe that as a diagnosis: if we don’t find a way to show concrete progress quickly, and use that to excite the executive team, our strategy is likely to fail. This transforms the thing preventing your strategy into a condition your strategy needs to address. Whenever you run into a reason why your strategy seems unlikely to work, or why strategy overall seems difficult, you’ve found an important piece of your diagnosis to include. There are never reasons why strategy simply cannot succeed, only diagnoses you’ve failed to recognize. For example, we knew in our work on Uber’s service provisioning strategy that we weren’t getting more headcount for the team, the product engineering team was going to continue growing rapidly, and that engineering leadership was unwilling to constrain how product engineering worked. Rather than preventing us from implementing a strategy, those components clarified what sort of approach could actually succeed. The role of self-awareness Every problem of today is partially rooted in the decisions of yesterday. If you’ve been with your organization for any duration at all, this means that you are directly or indirectly responsible for a portion of the problems that your diagnosis ought to recognize. This means that recognizing the impact of your prior actions in your diagnosis is a powerful demonstration of self-awareness. It also suggests that your next strategy’s success is rooted in your self-awareness about your prior choices. Don’t be afraid to recognize the failures in your past work. While changing your mind without new data is a sign of chaotic leadership, changing your mind with new data is a sign of thoughtful leadership. Summary Because diagnosis is the foundation of effective strategy, I’ve always found it the most intimidating phase of strategy work. While I think that’s a somewhat unavoidable reality, my hope is that this chapter has somewhat prepared you for that challenge. The four most important things to remember are simply: form your diagnosis before deciding how to solve it, try especially hard to capture perspectives you initially disagree with, supplement intuition with data where you can, and accept that sometimes you’re missing the data you need to fully understand. The last piece in particular, is why many good strategies never get shared, and the topic we’ll address in the next chapter on strategy refinement.
A Live, Interactive Course for Systems Engineers
I’m sitting in a small coffee shop in Brooklyn. I have a warm drink, and it’s just started to snow outside. I’m visiting New York to see Operation Mincemeat on Broadway – I was at the dress rehearsal yesterday, and I’ll be at the opening preview tonight. I’ve seen this show more times than I care to count, and I hope US theater-goers love it as much as Brits. The people who make the show will tell you that it’s about a bunch of misfits who thought they could do something ridiculous, who had the audacity to believe in something unlikely. That’s certainly one way to see it. The musical tells the true story of a group of British spies who tried to fool Hitler with a dead body, fake papers, and an outrageous plan that could easily have failed. Decades later, the show’s creators would mirror that same spirit of unlikely ambition. Four friends, armed with their creativity, determination, and a wardrobe full of hats, created a new musical in a small London theatre. And after a series of transfers, they’re about to open the show under the bright lights of Broadway. But when I watch the show, I see a story about friendship. It’s about how we need our friends to help us, to inspire us, to push us to be the best versions of ourselves. I see the swaggering leader who needs a team to help him truly achieve. The nervous scientist who stands up for himself with the support of his friends. The enthusiastic secretary who learns wisdom and resilience from her elder. And so, I suppose, it’s fitting that I’m not in New York on my own. I’m here with friends – dozens of wonderful people who I met through this ridiculous show. At first, I was just an audience member. I sat in my seat, I watched the show, and I laughed and cried with equal measure. After the show, I waited at stage door to thank the cast. Then I came to see the show a second time. And a third. And a fourth. After a few trips, I started to see familiar faces waiting with me at stage door. So before the cast came out, we started chatting. Those conversations became a Twitter community, then a Discord, then a WhatsApp. We swapped fan art, merch, and stories of our favourite moments. We went to other shows together, and we hung out outside the theatre. I spent New Year’s Eve with a few of these friends, sitting on somebody’s floor and laughing about a bowl of limes like it was the funniest thing in the world. And now we’re together in New York. Meeting this kind, funny, and creative group of people might seem as unlikely as the premise of Mincemeat itself. But I believed it was possible, and here we are. I feel so lucky to have met these people, to take this ridiculous trip, to share these precious days with them. I know what a privilege this is – the time, the money, the ability to say let’s do this and make it happen. How many people can gather a dozen friends for even a single evening, let alone a trip halfway round the world? You might think it’s silly to travel this far for a theatre show, especially one we’ve seen plenty of times in London. Some people would never see the same show twice, and most of us are comfortably into double or triple-figures. Whenever somebody asks why, I don’t have a good answer. Because it’s fun? Because it’s moving? Because I enjoy it? I feel the need to justify it, as if there’s some logical reason that will make all of this okay. But maybe I don’t have to. Maybe joy doesn’t need justification. A theatre show doesn’t happen without people who care. Neither does a friendship. So much of our culture tells us that it’s not cool to care. It’s better to be detached, dismissive, disinterested. Enthusiasm is cringe. Sincerity is weakness. I’ve certainly felt that pressure – the urge to play it cool, to pretend I’m above it all. To act as if I only enjoy something a “normal” amount. Well, fuck that. I don’t know where the drive to be detached comes from. Maybe it’s to protect ourselves, a way to guard against disappointment. Maybe it’s to seem sophisticated, as if having passions makes us childish or less mature. Or perhaps it’s about control – if we stay detached, we never have to depend on others, we never have to trust in something bigger than ourselves. Being detached means you can’t get hurt – but you’ll also miss out on so much joy. I’m a big fan of being a big fan of things. So many of the best things in my life have come from caring, from letting myself be involved, from finding people who are a big fan of the same things as me. If I pretended not to care, I wouldn’t have any of that. Caring – deeply, foolishly, vulnerably – is how I connect with people. My friends and I care about this show, we care about each other, and we care about our joy. That care and love for each other is what brought us together, and without it we wouldn’t be here in this city. I know this is a once-in-a-lifetime trip. So many stars had to align – for us to meet, for the show we love to be successful, for us to be able to travel together. But if we didn’t care, none of those stars would have aligned. I know so many other friends who would have loved to be here but can’t be, for all kinds of reasons. Their absence isn’t for lack of caring, and they want the show to do well whether or not they’re here. I know they care, and that’s the important thing. To butcher Tennyson: I think it’s better to care about something you cannot affect, than to care about nothing at all. In a world that’s full of cynicism and spite and hatred, I feel that now more than ever. I’d recommend you go to the show if you haven’t already, but that’s not really the point of this post. Maybe you’ve already seen Operation Mincemeat, and it wasn’t for you. Maybe you’re not a theatre kid. Maybe you aren’t into musicals, or history, or war stories. That’s okay. I don’t mind if you care about different things to me. (Imagine how boring the world would be if we all cared about the same things!) But I want you to care about something. I want you to find it, find people who care about it too, and hold on to them. Because right now, in this city, with these people, at this show? I’m so glad I did. And I hope you find that sort of happiness too. Some of the people who made this trip special. Photo by Chloe, and taken from her Twitter. Timing note: I wrote this on February 15th, but I delayed posting it because I didn’t want to highlight the fact I was away from home. [If the formatting of this post looks odd in your feed reader, visit the original article]
One of the biggest mistakes that new startup founders make is trying to get away from the customer-facing roles too early. Whether it's customer support or it's sales, it's an incredible advantage to have the founders doing that work directly, and for much longer than they find comfortable. The absolute worst thing you can do is hire a sales person or a customer service agent too early. You'll miss all the golden nuggets that customers throw at you for free when they're rejecting your pitch or complaining about the product. Seeing these reasons paraphrased or summarized destroy all the nutrients in their insights. You want that whole-grain feedback straight from the customers' mouth! When we launched Basecamp in 2004, Jason was doing all the customer service himself. And he kept doing it like that for three years!! By the time we hired our first customer service agent, Jason was doing 150 emails/day. The business was doing millions of dollars in ARR. And Basecamp got infinitely, better both as a market proposition and as a product, because Jason could funnel all that feedback into decisions and positioning. For a long time after that, we did "Everyone on Support". Frequently rotating programmers, designers, and founders through a day of answering emails directly to customers. The dividends of doing this were almost as high as having Jason run it all in the early years. We fixed an incredible number of minor niggles and annoying bugs because programmers found it easier to solve the problem than to apologize for why it was there. It's not easy doing this! Customers often offer their valuable insights wrapped in rude language, unreasonable demands, and bad suggestions. That's why many founders quit the business of dealing with them at the first opportunity. That's why few companies ever do "Everyone On Support". That's why there's such eagerness to reduce support to an AI-only interaction. But quitting dealing with customers early, not just in support but also in sales, is an incredible handicap for any startup. You don't have to do everything that every customer demands of you, but you should certainly listen to them. And you can't listen well if the sound is being muffled by early layers of indirection.
