SEARCH WEB FOR A PLANET: THE GOOGLE CLUSTER ARCHITECTURE
AMENABLE TO EXTENSIVE PARALLELIZATION, GOOGLES WEB SEARCH
APPLICATION LETS DIFFERENT QUERIES RUN ON DIFFERENT PROCESSORS AND, BY PARTITIONING THE OVERALL INDEX, ALSO LETS A SINGLE QUERY USE MULTIPLE PROCESSORS. TO HANDLE THIS WORKLOAD, GOOGLES ARCHITECTURE FEATURES CLUSTERS OF MORE THAN 15,000 COMMODITYCLASS PCS WITH FAULT-TOLERANT SOFTWARE. THIS ARCHITECTURE ACHIEVES SUPERIOR PERFORMANCE AT A FRACTION OF THE COST OF A SYSTEM BUILT FROM FEWER, BUT MORE EXPENSIVE, HIGH-END SERVERS.
Luiz Andr Barroso Jeffrey Dean Urs Hlzle Google
Few Web services require as much computation per request as search engines. On average, a single query on Google reads hundreds of megabytes of data and consumes tens of billions of CPU cycles. Supporting a peak request stream of thousands of queries per second requires an infrastructure comparable in size to that of the largest supercomputer installations. Combining more than 15,000 commodity-class PCs with fault-tolerant software creates a solution that is more cost-effective than a comparable system built out of a smaller number of high-end servers. Here we present the architecture of the Google cluster, and discuss the most important factors that inuence its design: energy efciency and price-performance ratio. Energy efciency is key at our scale of operation, as power consumption and cooling issues become signicant operational factors, taxing the lim-
its of available data center power densities. Our application affords easy parallelization: Different queries can run on different processors, and the overall index is partitioned so that a single query can use multiple processors. Consequently, peak processor performance is less important than its price/ performance. As such, Google is an example of a throughput-oriented workload, and should benefit from processor architectures that offer more on-chip parallelism, such as simultaneous multithreading or on-chip multiprocessors.
Google architecture overview
Googles software architecture arises from two basic insights. First, we provide reliability in software rather than in server-class hardware, so we can use commodity PCs to build a high-end computing cluster at a low-end
22
Published by the IEEE Computer Society
0272-1732/03/$17.00 2003 IEEE
price. Second, we tailor the design for best aggregate request throughput, not peak server response time, since we can manage response times by parallelizing individual requests. We believe that the best price/performance tradeoff for our applications comes from fashioning a reliable computing infrastructure from clusters of unreliable commodity PCs. We provide reliability in our environment at the software level, by replicating services across many different machines and automatically detecting and handling failures. This softwarebased reliability encompasses many different areas and involves all parts of our system design. Examining the control flow in handling a query provides insight into the highlevel structure of the query-serving system, as well as insight into reliability considerations.
Google Web server
Spell checker
Ad server
Index servers
Document servers
Figure 1. Google query-serving architecture.
Serving a Google query
When a user enters a query to Google (such as www.google.com/search?q=ieee+society), the users browser rst performs a domain name system (DNS) lookup to map www.google.com to a particular IP address. To provide sufcient capacity to handle query trafc, our service consists of multiple clusters distributed worldwide. Each cluster has around a few thousand machines, and the geographically distributed setup protects us against catastrophic data center failures (like those arising from earthquakes and large-scale power failures). A DNS-based load-balancing system selects a cluster by accounting for the users geographic proximity to each physical cluster. The load-balancing system minimizes round-trip time for the users request, while also considering the available capacity at the various clusters. The users browser then sends a hypertext transport protocol (HTTP) request to one of these clusters, and thereafter, the processing of that query is entirely local to that cluster. A hardware-based load balancer in each cluster monitors the available set of Google Web servers (GWSs) and performs local load balancing of requests across a set of them. After receiving a query, a GWS machine coordinates the query execution and formats the results into a Hypertext Markup Language (HTML) response to the users browser. Figure 1 illustrates these steps. Query execution consists of two major phases.1 In the first phase, the index servers
consult an inverted index that maps each query word to a matching list of documents (the hit list). The index servers then determine a set of relevant documents by intersecting the hit lists of the individual query words, and they compute a relevance score for each document. This relevance score determines the order of results on the output page. The search process is challenging because of the large amount of data: The raw documents comprise several tens of terabytes of uncompressed data, and the inverted index resulting from this raw data is itself many terabytes of data. Fortunately, the search is highly parallelizable by dividing the index into pieces (index shards), each having a randomly chosen subset of documents from the full index. A pool of machines serves requests for each shard, and the overall index cluster contains one pool for each shard. Each request chooses a machine within a pool using an intermediate load balancerin other words, each query goes to one machine (or a subset of machines) assigned to each shard. If a shards replica goes down, the load balancer will avoid using it for queries, and other components of our cluster-management system will try to revive it or eventually replace it with another machine. During the downtime, the system capacity is reduced in proportion to the total fraction of capacity that this machine represented. However, service remains uninterrupted, and all parts of the index remain available. The nal result of this rst phase of query execution is an ordered list of document identiers (docids). As Figure 1 shows, the second
MARCHAPRIL 2003
23
GOOGLE SEARCH ARCHITECTURE
phase involves taking this list of docids and computing the actual title and uniform resource locator of these documents, along with a query-specific document summary. Document servers (docservers) handle this job, fetching each document from disk to extract the title and the keyword-in-context snippet. As with the index lookup phase, the strategy is to partition the processing of all documents by randomly distributing documents into smaller shards having multiple server replicas responsible for handling each shard, and routing requests through a load balancer. The docserver cluster must have access to an online, low-latency copy of the entire Web. In fact, because of the replication required for performance and availability, Google stores dozens of copies of the Web across its clusters. In addition to the indexing and documentserving phases, a GWS also initiates several other ancillary tasks upon receiving a query, such as sending the query to a spell-checking system and to an ad-serving system to generate relevant advertisements (if any). When all phases are complete, a GWS generates the appropriate HTML for the output page and returns it to the users browser.
allelizing the search over many machines, we reduce the average latency necessary to answer a query, dividing the total computation across more CPUs and disks. Because individual shards dont need to communicate with each other, the resulting speedup is nearly linear. In other words, the CPU speed of the individual index servers does not directly influence the searchs overall performance, because we can increase the number of shards to accommodate slower CPUs, and vice versa. Consequently, our hardware selection process focuses on machines that offer an excellent request throughput for our application, rather than machines that offer the highest singlethread performance. In summary, Google clusters follow three key design principles: Software reliability. We eschew fault-tolerant hardware features such as redundant power supplies, a redundant array of inexpensive disks (RAID), and highquality components, instead focusing on tolerating failures in software. Use replication for better request throughput and availability. Because machines are inherently unreliable, we replicate each of our internal services across many machines. Because we already replicate services across multiple machines to obtain sufficient capacity, this type of fault tolerance almost comes for free. Price/performance beats peak performance. We purchase the CPU generation that currently gives the best performance per unit price, not the CPUs that give the best absolute performance. Using commodity PCs reduces the cost of computation. As a result, we can afford to use more computational resources per query, employ more expensive techniques in our ranking algorithm, or search a larger index of documents.
Using replication for capacity and fault-tolerance
We have structured our system so that most accesses to the index and other data structures involved in answering a query are read-only: Updates are relatively infrequent, and we can often perform them safely by diverting queries away from a service replica during an update. This principle sidesteps many of the consistency issues that typically arise in using a general-purpose database. We also aggressively exploit the very large amounts of inherent parallelism in the application: For example, we transform the lookup of matching documents in a large index into many lookups for matching documents in a set of smaller indices, followed by a relatively inexpensive merging step. Similarly, we divide the query stream into multiple streams, each handled by a cluster. Adding machines to each pool increases serving capacity, and adding shards accommodates index growth. By par-
Leveraging commodity parts
Googles racks consist of 40 to 80 x86-based servers mounted on either side of a custom made rack (each side of the rack contains twenty 20u or forty 1u servers). Our focus on price/performance favors servers that resemble mid-range desktop PCs in terms of their components, except for the choice of large disk
24
IEEE MICRO
drives. Several CPU generations are in active service, ranging from single-processor 533MHz Intel-Celeron-based servers to dual 1.4GHz Intel Pentium III servers. Each server contains one or more integrated drive electronics (IDE) drives, each holding 80 Gbytes. Index servers typically have less disk space than document servers because the former have a more CPU-intensive workload. The servers on each side of a rack interconnect via a 100-Mbps Ethernet switch that has one or two gigabit uplinks to a core gigabit switch that connects all racks together. Our ultimate selection criterion is cost per query, expressed as the sum of capital expense (with depreciation) and operating costs (hosting, system administration, and repairs) divided by performance. Realistically, a server will not last beyond two or three years, because of its disparity in performance when compared to newer machines. Machines older than three years are so much slower than current-generation machines that it is difficult to achieve proper load distribution and conguration in clusters containing both types. Given the relatively short amortization period, the equipment cost gures prominently in the overall cost equation. Because Google servers are custom made, well use pricing information for comparable PC-based server racks for illustration. For example, in late 2002 a rack of 88 dual-CPU 2-GHz Intel Xeon servers with 2 Gbytes of RAM and an 80-Gbyte hard disk was offered on RackSaver.com for around $278,000. This gure translates into a monthly capital cost of $7,700 per rack over three years. Personnel and hosting costs are the remaining major contributors to overall cost. The relative importance of equipment cost makes traditional server solutions less appealing for our problem because they increase performance but decrease the price/performance. For example, four-processor motherboards are expensive, and because our application parallelizes very well, such a motherboard doesnt recoup its additional cost with better performance. Similarly, although SCSI disks are faster and more reliable, they typically cost two or three times as much as an equal-capacity IDE drive. The cost advantages of using inexpensive, PC-based clusters over high-end multi-
processor servers can be quite substantial, at least for a highly parallelizable application like ours. The example $278,000 rack contains 176 2-GHz Xeon CPUs, 176 Gbytes of RAM, and 7 Tbytes of disk space. In comparison, a typical x86-based server contains eight 2-GHz Xeon CPUs, 64 Gbytes of RAM, and 8 Tbytes of disk space; it costs about $758,000.2 In other words, the multiprocessor server is about three times more expensive but has 22 times fewer CPUs, three times less RAM, and slightly more disk space. Much of the cost difference derives from the much higher interconnect bandwidth and reliability of a high-end server, but again, Googles highly redundant does architecture not rely on either of these attributes. Operating thousands of mid-range PCs instead of a few high-end multiprocessor servers incurs signicant system administration and repair costs. However, for a relatively homogenous application like Google, where most servers run one of very few applications, these costs are manageable. Assuming tools to install and upgrade software on groups of machines are available, the time and cost to maintain 1,000 servers isnt much more than the cost of maintaining 100 servers because all machines have identical congurations. Similarly, the cost of monitoring a cluster using a scalable application-monitoring system does not increase greatly with cluster size. Furthermore, we can keep repair costs reasonably low by batching repairs and ensuring that we can easily swap out components with the highest failure rates, such as disks and power supplies.
The power problem
Even without special, high-density packaging, power consumption and cooling issues can become challenging. A mid-range server with dual 1.4-GHz Pentium III processors draws about 90 W of DC power under load: roughly 55 W for the two CPUs, 10 W for a disk drive, and 25 W to power DRAM and the motherboard. With a typical efciency of about 75 percent for an ATX power supply, this translates into 120 W of AC power per server, or roughly 10 kW per rack. A rack comfortably ts in 25 ft2 of space, resulting in a power density of 400 W/ft2. With higher-end processors, the power density of a rack can exceed 700 W/ft2.
MARCHAPRIL 2003
25
GOOGLE SEARCH ARCHITECTURE
Table 1. Instruction-level measurements on the index server.
Characteristic Cycles per instruction Ratios (percentage) Branch mispredict Level 1 instruction miss* Level 1 data miss* Level 2 miss* Instruction TLB miss* Data TLB miss* Value 1.1 5.0 0.4 0.7 0.3 0.04 0.7
Hardware-level application characteristics
Examining various architectural characteristics of our application helps illustrate which hardware platforms will provide the best price/performance for our query-serving system. Well concentrate on the characteristics of the index server, the component of our infrastructure whose price/performance most heavily impacts overall price/performance. The main activity in the index server consists of decoding compressed information in the inverted index and nding matches against a set of documents that could satisfy a query. Table 1 shows some basic instruction-level measurements of the index server program running on a 1-GHz dualprocessor Pentium III system. The application has a moderately high CPI, considering that the Pentium III is capable of issuing three instructions per cycle. We expect such behavior, considering that the application traverses dynamic data structures and that control ow is data dependent, creating a significant number of difficult-to-predict branches. In fact, the same workload running on the newer Pentium 4 processor exhibits nearly twice the CPI and approximately the same branch prediction performance, even though the Pentium 4 can issue more instructions concurrently and has superior branch prediction logic. In essence, there isnt that much exploitable instruction-level parallelism (ILP) in the workload. Our measurements suggest that the level of aggressive out-oforder, speculative execution present in modern processors is already beyond the point of diminishing performance returns for such programs. A more protable way to exploit parallelism for applications such as the index server is to leverage the trivially parallelizable computation. Processing each query shares mostly readonly data with the rest of the system, and constitutes a work unit that requires little communication. We already take advantage of that at the cluster level by deploying large numbers of inexpensive nodes, rather than fewer highend ones. Exploiting such abundant threadlevel parallelism at the microarchitecture level appears equally promising. Both simultaneous multithreading (SMT) and chip multiprocessor (CMP) architectures target thread-level parallelism and should improve the performance of many of our servers. Some early
* Cache and TLB ratios are per instructions retired.
Unfortunately, the typical power density for commercial data centers lies between 70 and 150 W/ft2, much lower than that required for PC clusters. As a result, even low-tech PC clusters using relatively straightforward packaging need special cooling or additional space to bring down power density to that which is tolerable in typical data centers. Thus, packing even more servers into a rack could be of limited practical use for large-scale deployment as long as such racks reside in standard data centers. This situation leads to the question of whether it is possible to reduce the power usage per server. Reduced-power servers are attractive for large-scale clusters, but you must keep some caveats in mind. First, reduced power is desirable, but, for our application, it must come without a corresponding performance penalty: What counts is watts per unit of performance, not watts alone. Second, the lower-power server must not be considerably more expensive, because the cost of depreciation typically outweighs the cost of power. The earlier-mentioned 10 kW rack consumes about 10 MW-h of power per month (including cooling overhead). Even at a generous 15 cents per kilowatt-hour (half for the actual power, half to amortize uninterruptible power supply [UPS] and power distribution equipment), power and cooling cost only $1,500 per month. Such a cost is small in comparison to the depreciation cost of $7,700 per month. Thus, low-power servers must not be more expensive than regular servers to have an overall cost advantage in our setup.
26
IEEE MICRO
experiments with a dual-context (SMT) Intel Xeon processor show more than a 30 percent performance improvement over a single-context setup. This speedup is at the upper bound of improvements reported by Intel for their SMT implementation.3 We believe that the potential for CMP systems is even greater. CMP designs, such as Hydra4 and Piranha,5 seem especially promising. In these designs, multiple (four to eight) simpler, in-order, short-pipeline cores replace a complex high-performance core. The penalties of in-order execution should be minor given how little ILP our application yields, and shorter pipelines would reduce or eliminate branch mispredict penalties. The available thread-level parallelism should allow near-linear speedup with the number of cores, and a shared L2 cache of reasonable size would speed up interprocessor communication.
Large-scale multiprocessing
As mentioned earlier, our infrastructure consists of a massively large cluster of inexpensive desktop-class machines, as opposed to a smaller number of large-scale sharedmemory machines. Large shared-memory machines are most useful when the computation-to-communication ratio is low; communication patterns or data partitioning are dynamic or hard to predict; or when total cost of ownership dwarfs hardware costs (due to management overhead and software licensing prices). In those situations they justify their high price tags. At Google, none of these requirements apply, because we partition index data and computation to minimize communication and evenly balance the load across servers. We also produce all our software in-house, and minimize system management overhead through extensive automation and monitoring, which makes hardware costs a signicant fraction of the total system operating expenses. Moreover, large-scale shared-memory machines still do not handle individual hardware component or software failures gracefully, with most fault types causing a full system crash. By deploying many small multiprocessors, we contain the effect of faults to smaller pieces of the system. Overall, a cluster solution ts the performance and availability requirements of our service at significantly lower costs. t rst sight, it might appear that there are few applications that share Googles ...